Can you share the story behind the founding of Materials Nexus? What inspired the creation of the company and its focus on AI-driven materials discovery?

 

Ultimately, the limit of what can be built is the materials used to build it; that was my motivation to study materials science. During my time at University of Cambridge, working with my co-founder Robert Forrest, the desire to make our research go faster inspired our pivot towards the development of machine learning algorithms. This became the foundation of Materials Nexus’ technology.

 

It was clear that this research could have a positive impact in the world and its adoption needed to be accelerated. In the same way, the performance of products is limited by materials, so is our progress towards net-zero. This is what inspired us to found the business.

 

Can you explain how AI is transforming the process of materials discovery, particularly in the context of Materials Nexus?

 

In the same way AI impacted the drug discovery process, it is also fundamentally changing materials discovery; transforming what is typically a trial-and-error-based approach to an intent-based design process. But unlike pharmaceutical research, there is the added complexity and a wider search space across the entire periodic table. At Materials Nexus, we’re looking at the entire length-scale, from quantum level to bulk – this means that we are not only leveraging quantum mechanics for composition prediction but also modelling processing and synthesis techniques. This allows us to not only identify, but also physically produce high-performance materials accurately, in a matter of months rather than decades, significantly speeding up the R&D process.

 

What are the key benefits of using AI over traditional trial-and-error methods in developing new materials?

 

Using AI for materials discovery offers several benefits: speed, cost-efficiency, and sustainability being key. Our AI-driven platform can analyze vast datasets and predict material properties accurately, all before setting foot in a lab, making the process cost-effective and less wasteful, as it minimizes the need for expensive and resource-intensive experiments. This also means processes that usually take days in a lab could be done in hours on our platform.

 

This ultimately unlocks a new set of opportunities with targeted material “design” vs. discovery. It is possible to incorporate any data set or material parameter, such as CO2 emissions, cost, or weight, and search for compositions to match those specific needs, flipping the “discovery” process on its head.

 

What role do AI and machine learning play in reducing the environmental impact of material production?

 

Leveraging AI and machine learning unlocks a vast new set of material opportunities through the discovery phase. At the production level, the impact of this is two-fold; first is the elemental composition of the materials themselves, second is the materials’ processing conditions. AI materials discovery can either exclude specific elements that have a high environmental cost (e.g. rare earth elements) or reduce their compositional percentage. It can also be used to look at processing techniques (e.g. the temperature, pressure or even purity of ore) required to make the material and identify low-energy methods. These two aspects can have a significant impact on the primary emissions of material production. However, it is important to note that environmental impact goes beyond production alone. The application of superior materials, both high performance or cheaper, can have a hugely positive secondary environmental impact by making sustainable technologies more accessible (e.g. cheaper EVs), more efficient (e.g. better computer chips for AI), and less toxic in their end-of-life disposal (e.g. replacing hydrofluorocarbons).

 

Can AI potentially replace rare earth metals in other applications beyond magnets?

 

AI powered material discovery has the potential to identify and develop alternative materials for a vast range of applications beyond magnets. In this instance the aim was to find an alternative magnet composition that removed rare-earth elements, but our machine learning search algorithms are built to be applied to any material class. This means that we are building a universal materials design platform.

Viewed By: 77355933